🚀 go-pugleaf

jerry@msi.com (Jerry Shekhel) writes:

>Kendall Bennett (rcskb@minyos.xx.rmit.OZ.AU) wrote:
>:
>: We are talking about games here - windows is not a games engine. Windows is
>: great for software like word processors and spreadsheets that need to have
>: a nice looking and reasonably responsive user interface, but it is hardly
>: workable if you are looking for real time graphics, especially 3d.
>:

>Quite wrong.  We now have graphics accelerators for Windows -- ATI Mach32,
>S3, XGA, etc.  The beauty of it is that with Windows, we have a single high-
>level API that works on all of these.

Ahh, but for fast 3D we need high performance convex polygon scan conversion
code - none of the accelerator chips are going to provide you with this. 
Windows only has a complex polygon scan converter, which is just too slow
for real 3D - when OpenGL support comes to PC user interfaces we might
just have an option.

Basically the GUI accelerators simply speed up rectangle fill's, line 
drawing and BitBlt'ing...

>:
>: Ever heard of double or triple buffering in Windows?
>:

>Sure.  My game, Chomp, uses two offscreen buffers.  It's not synchronized
>to the display, and screen updates have to be done using blitting instead
>of hardware page-flipping, but it does use double buffering.

Hah. This is less than satifactory for 3D, especially when the bitmap are
large. An 800x600 256 color bitmap takes up about 512k of memory.
Transferring all this to the display memory takes a while, and you cannot
ensure that it will be done during a vertical retrace.

Once again, buying a GUI accelerator card is not going to solve the problem.
The GUI accelerators currently available still have to transfer the data
across the system BUS to do a memory to screen BitBlt (they are _very_
fast a moving stuff around the screen, but this is useless for this
application). Cards like the S3 based on the 8514/A move all the data
through an I/O port, or you can simply use the card like any dumb frame
buffer SVGA (like an ET4000!).

Newer cards like the XGA, and I believe the Mach 32 can do bus mastering
to move the data a whole lot faster, but unless the card is a local bus
implementation, MCA or EISA it will still be slow.

+------------------------------------------+-------------------------------+
| Kendall Bennett                          | Internet:                     |
| RMIT Advanced Computer Graphics Centre   | kjb@citri.edu.au              |
| CITRI Building, 723 Swanston Street      | rcskb@minyos.xx.rmit.oz.au    |
| Carlton Victoria 3053 AUSTRALIA.         |                               | 
+------------------------------------------+-------------------------------+